Quantification of Drift in Battery Impedance Spectroscopy Due to State of Charge, Current Amplitude, and Rest Time

Electrochemical impedance spectroscopy (EIS) is a powerful, non-destructive probe for lithium-ion battery state-of-health (SOH) estimation, but its reliability is highly sensitive to operating-point choices. This paper quantifies how state of charge (SOC) levels, AC excitation signal amplitudes, and pre-test rest time propagate variance and bias into the impedance spectrum using EIS data collected form 240 experiments spanning four cells and 60 different operating points. The analysis shows that mid-range SOC and long rest durations yield the most stable spectra. Moreover, a clear noise–bias trade-off with excitation amplitude is established in this work, i.e., small perturbations inflate estimator variance, whereas large perturbations bias the ohmic resistance; a mid-range amplitude provides a robust operating point for stable SOH predictions. To reduce the uncertainties due to operating point variations, a novel impedance feature called thed-Ohm metricis proposed. The d-Ohm metric demonstrates lower sensitivity to operating conditions and reduced cell-to-cell variability. The proposed metric improves reproducibility (across cells and runs) and shows potential to support uncertainty-aware SOH estimation with confidence-calibrated outputs. Collectively, these results offer practical prescriptions for EIS test design and strengthen the robustness of both model-based and data-driven SOH estimation.